U.S. patent number 7,286,593 [Application Number 10/195,234] was granted by the patent office on 2007-10-23 for system and method for channel estimation in a radio frequency receiver.
This patent grant is currently assigned to National Semiconductor Corporation. Invention is credited to Debarag N. Banerjee.
United States Patent |
7,286,593 |
Banerjee |
October 23, 2007 |
System and method for channel estimation in a radio frequency
receiver
Abstract
A channel estimator for determining channel weighting
coefficients for a finger of the RAKE receiver. The channel
estimator comprises: 1) a first correlator for receiving a first
pilot channel signal and correlating the first pilot channel signal
with a first pilot channel chip pattern associated with the first
pilot channel signal to produce an output comprising a first pilot
channel symbol sequence; and 2) a first channel estimation filter
capable of receiving the first pilot channel symbol sequence and
generating first channel weighting coefficients. The first channel
estimation filter minimizes the mean squared error of the channel
estimate in the first channel weighting coefficients caused by
additive noise and Doppler effects, wherein the mean squared error
is minimized across a range of Doppler frequencies from 0 Hz up to
a predetermined maximum Doppler frequency.
Inventors: |
Banerjee; Debarag N.
(Sunnyvale, CA) |
Assignee: |
National Semiconductor
Corporation (Santa Clara, CA)
|
Family
ID: |
38607084 |
Appl.
No.: |
10/195,234 |
Filed: |
July 15, 2002 |
Current U.S.
Class: |
375/148; 375/142;
375/152; 375/E1.032 |
Current CPC
Class: |
H04B
1/71055 (20130101); H04B 1/709 (20130101); H04B
1/712 (20130101) |
Current International
Class: |
H04B
1/00 (20060101) |
Field of
Search: |
;375/142,143,147,148,150,152 |
References Cited
[Referenced By]
U.S. Patent Documents
Other References
Ji-Woong Choi and Yong-Hwan Lee, "An adaptive channel estimator in
pilot channel based DS-CDMA systems", IEEE 55th Vehicular
Technology Conference, 2002. VTC Spring 2002, vol. 3, May 6-9, 2002
pp. 1429-1433 vol. 3. cited by examiner .
Karam, "Digital filtering", chapter 11 of the book "Digital Signal
Processing Handbook" edited by Douglas B. Williams and Vijay K.
Madisetti, CRC press 1997 pp. 11-32 to 11-35. cited by examiner
.
Ji-Woong Choi and Yong-Hwan Lee, "Adaptive channel estimation in
DS-CDMA downlink systems", The 13th IEEE International Symposium on
Personal, Indoor and Mobile Radio Communications, 2002. vol. 3,
Sep. 15-18, 2002 pp. 1432-1436 vol. 3. cited by examiner .
Sakamoto, "Adaptive channel estimation with velocity estimator for
W-CDMA receiver", IEEE 51st Vehicular Technology Conference
Proceedings, 2000, VTC 2000-Spring Tokyo. 2000 vol. 3, May 15-18,
2000 pp. 2024-2028 vol. 3. cited by examiner .
Karam, "Digital filtering", chapter 11 of the book "Digital Signal
Processing Handbook" edited by Douglas B. Williams and Vijay K.
Madisetti, CRC press 1997 pp. 11-40 to 11-42. cited by examiner
.
Mitra, "Handbook for digital signal processing", 1993 pp. 425-427.
cited by examiner .
Choi, "Design of channel estimation filter in DS-CDMA uplink
systems," IEEE ISPACS'00, pp. 1011-1016, Nov. 2000. cited by
examiner .
Ji-Woong Choi et al., "Adaptive Channel Estimation in DS-CDMA
Downlink Systems," 2002 IEEE, pp. 1432-1436. cited by other .
Ji-Woong Choi et al., "An Adaptive Channel Estimator in Pilot
Channel Based DS-CDMA Systems," 2002 IEEE, pp. 1429-1433. cited by
other .
Lina J. Karam et al., "Digital Filtering," The Digital Signal
Processing Handbook, 1998 by CRC Press LLC, pp. 11-1-11-35. cited
by other.
|
Primary Examiner: Ghebretinsae; Temesghen
Assistant Examiner: Torres; Juan Alberto
Claims
What is claimed is:
1. A channel estimator for determining channel weighting
coefficients for a finger of a RAKE receiver, said channel
estimator comprising: a correlator configured to receive a pilot
channel signal and to correlate said pilot channel signal with a
pilot channel chip pattern to produce a pilot channel symbol
sequence; and a channel estimation filter configured to receive
said pilot channel symbol sequence and to generate channel
weighting coefficients, wherein said channel estimation filter is
configured to minimize a mean squared error of a channel estimate
in said channel weighting coefficients caused by at least one of:
additive noise and Doppler effects, wherein said mean squared error
is minimized across a range of Doppler frequencies from 0 Hz up to
a specified maximum Doppler frequency; wherein the channel
estimation filter comprises a first filter, a second filter, and an
adder, the adder configured to sum outputs of the first and second
filters, and inputs of the second filter comprising outputs of the
adder.
2. The channel estimator as set forth in claim 1 wherein said
channel estimation filter is configured to minimize an average mean
squared error over the entire Doppler range assuming that a mobile
station velocity is uniformly distributed.
3. The channel estimator as set forth in claim 2 wherein said
channel estimation filter is configured to minimize the average
mean squared error over the entire Doppler range assuming a power
spectral density of the received pilot channel signal follows a
Jake's spectrum.
4. The channel estimator as set forth in claim 1 wherein said first
filter comprises a 3-tap finite impulse response (FIR) filter and
the second filter comprises a single pole infinite impulse response
(IIR) filter.
5. The channel estimator as set forth in claim 4 wherein said
single pole infinite impulse response (IIR) filter is a Butterworth
filter having a maximally flat passband.
6. The channel estimator as set forth in claim 1 wherein one of the
first and second filters in said channel estimation filter
comprises a single pole infinite impulse response (IIR) filter
which is a Butterworth filter having a maximally flat passband.
7. The channel estimator as set forth in claim 1 further
comprising: a second correlator configured to receive a second
pilot channel signal and to correlate said second pilot channel
signal with a second pilot channel chip pattern to produce a second
pilot channel symbol sequence; and a second channel estimation
filter configured to receive said second pilot channel symbol
sequence and to generate second channel weighting coefficients,
wherein said second channel estimation filter is configured to
minimize a second mean squared error of a second channel estimate
in said second channel weighting coefficients caused by at least
one of: the additive noise and the Doppler effects, wherein said
second mean squared error is minimized across the range of Doppler
frequencies.
8. The channel estimator as set forth in claim 7 wherein said
channel estimation filter is configured to minimize a second
average mean squared error over the entire Doppler range assuming
that a mobile station velocity is uniformly distributed.
9. The channel estimator as set forth in claim 8 wherein said
channel estimation filter is configured to minimize the second
average mean squared error over the entire Doppler range assuming a
power spectral density of the received second pilot channel signal
follows a Jake's spectrum.
10. The channel estimator as set forth in claim 7 wherein said
second channel estimation filter comprises a cascade of a 3-tap
finite impulse response (FIR) filter and a single pole infinite
impulse response (IIR) filter.
11. The channel estimator as set forth in claim 10 wherein said
single pole infinite impulse response (IIR) filter is a Butterworth
filter having a maximally flat passband.
12. The channel estimator as set forth in claim 7 wherein said
second channel estimation filter comprises a single pole infinite
impulse response (IIR) filter which is a Butterworth filter having
a maximally flat passband.
13. The channel estimator as set forth in claim 7, further
comprising: a first multiplier configured to multiply the channel
weighting coefficients and the pilot channel symbol sequence to
produce first weighted outputs; a second multiplier configured to
multiply the second channel weighting coefficients and the second
pilot channel symbol sequence to produce second weighted outputs;
and a combiner configured to sum the first and second weighted
outputs to produce an output signal.
14. The channel estimator as set forth in claim 1, wherein the
range of Doppler frequencies comprises a range from 0 Hz to 500
Hz.
15. A RAKE receiver comprising: a radio frequency (RF) front-end
configured to receive an incoming RF signal and to down-convert and
digitize said RF signal to a baseband or intermediate signal
comprising a sequence of digital samples; multiple finger elements,
each of said finger elements configured to delay and correlate a
received copy of said sequence of digital samples to thereby
produce a correlated output, wherein each finger element is
configured to multiply said correlated output by a corresponding
channel weighting coefficient associated with said finger element;
and a channel estimator for determining channel weighting
coefficients for said finger elements, said channel estimator
comprising: a correlator configured to receive a pilot channel
signal and to correlate said pilot channel signal with a pilot
channel chip pattern to produce a pilot channel symbol sequence;
and a channel estimation filter configured to receive said pilot
channel symbol sequence and to generate channel weighting
coefficients, wherein said channel estimation filter is configured
to minimize a mean squared error in said channel weighting
coefficients caused by at least one of: additive noise and Doppler
effects, wherein said mean squared error is minimized across a
range of Doppler frequencies from 0 Hz up to a specified maximum
Doppler frequency; wherein the channel estimation filter comprises
a first filter, a second filter, and an adder, the adder configured
to sum outputs of the first and second filters, and inputs of the
second filter comprising outputs of the adder.
16. The RAKE receiver as set forth in claim 15 wherein said channel
estimation filter is configured to minimize an average mean squared
error over the entire Doppler range assuming that a mobile station
velocity is uniformly distributed.
17. The RAKE receiver as set forth in claim 16 wherein said channel
estimation filter is configured to minimize the average mean
squared error over the entire Doppler range assuming a power
spectral density of the received pilot channel signal follows a
Jake's spectrum.
18. The RAKE receiver as set forth in claim 15 wherein said first
filter comprises a 3-tap finite impulse response (FIR) filter and
the second filter comprises a single pole infinite impulse response
(IIR) filter.
19. The RAKE receiver as set forth in claim 18 wherein said single
pole infinite impulse response (IIR) filter is a Butterworth filter
having a maximally flat passband.
20. The RAKE receiver as set forth in claim 15 wherein one of the
first and second filters in said channel estimation filter
comprises a single pole infinite impulse response (IIR) filter
which is a Butterworth filter having a maximally flat passband.
21. The RAKE receiver as set forth in claim 15 further comprising:
a second correlator configured to receive a second pilot channel
signal and to correlate said second pilot channel signal with a
second pilot channel chip pattern to produce a second pilot channel
symbol sequence; and a second channel estimation filter configured
to receive said second pilot channel symbol sequence and to
generate second channel weighting coefficients, wherein said second
channel estimation filter is configured to minimize a second mean
squared error of a second channel estimate in said second channel
weighting coefficients caused by at least one of: the additive
noise and the Doppler effects, wherein said second mean squared
error is minimized across the range of Doppler frequencies.
22. The RAKE receiver as set forth in claim 21 wherein said channel
estimation filter is configured to minimize a second average mean
squared error over the entire Doppler range assuming that a mobile
station velocity is uniformly distributed.
23. The RAKE receiver as set forth in claim 22 wherein said channel
estimation filter is configured to minimize the second average mean
squared error over the entire Doppler range assuming a power
spectral density of the received second pilot channel signal
follows a Jake's spectrum.
24. The RAKE receiver as set forth in claim 21 wherein said second
channel estimation filter comprises a cascade of a 3-tap finite
impulse response (FIR) filter and a single pole infinite impulse
response (IIR) filter.
25. The RAKE receiver as set forth in claim 24 wherein said single
pole infinite impulse response (IIR) filter is a Butterworth filter
having a maximally flat passband.
26. The RAKE receiver as set forth in claim 21 wherein said second
channel estimation filter comprises a single pole infinite impulse
response (IIR) filter which is a Butterworth filter having a
maximally flat passband.
27. A method for determining channel weighting coefficients for a
finger of a RAKE receiver, the method comprising the steps of:
receiving a pilot channel signal from a base station; correlating
the pilot channel signal with a pilot channel chip pattern to
produce a pilot channel symbol sequence; and generating from the
pilot channel symbol sequence channel weighting coefficients and
minimizing a mean squared error in the channel weighting
coefficients caused by at least one of: additive noise and Doppler
effects using a first filter, a second filter, and an adder,
wherein the adder sums outputs of the first and second filters,
wherein inputs of the second filter comprise outputs of the adder,
and wherein the mean squared error is minimized across a range of
Doppler frequencies from 0 Hz up to a specified maximum Doppler
frequency.
28. The method as set forth in claim 27 wherein the step of
minimizing minimizes an average mean squared error over the entire
Doppler range assuming that a mobile station velocity is uniformly
distributed.
29. The method as set forth in claim 28 wherein the step of
minimizing minimizes the average mean squared error over the entire
Doppler range assuming a power spectral density of the received
pilot channel signal follows a Jake's spectrum.
Description
TECHNICAL FIELD OF THE INVENTION
The present invention relates generally to wireless receivers and,
more particularly, to an apparatus and a related method in a
wireless receiver that performs channel estimation using minimum
mean squared error (MMSE) across a range of Doppler
frequencies.
BACKGROUND OF THE INVENTION
Business and consumers use a wide array of wireless devices,
including cell phones, wireless local area network (LAN) cards,
global positioning system (GPS) devices, electronic organizers
equipped with wireless modems, and the like. The increased demand
for wireless communication devices has created a corresponding
demand for technical improvements to such devices. Generally
speaking, wireless system designers attempt to minimize the cost of
conventional radio receivers while improving the performance of
such devices. Performance improvements include, among other things,
lower power consumption, greater range, increased receiver
sensitivity, lower bit error rates (BER), higher transmission
rates, and the like.
Signal fading due to variations in channel characteristics is a
major factor limiting the performance of modern mobile wireless
communication systems. To compensate for signal fading, many modern
code division multiple access (CDMA) networks use diversity
techniques to transmit multiples copies of a signal over a channel
to a mobile station. In the mobile station, a RAKE receiver uses
multiple baseband correlators to individually process several
signal multipath components. The correlator outputs are then
combined to achieve improved performance.
However, a RAKE receiver assumes that the channel variations over
time are relatively slow. This may not be the case if the mobile
station is moving relatively quickly. Channel variations due to
Doppler effects caused by the relative motion of the base station
transmitter and the mobile station receiver often become
significant. To correct this, receivers have been developed that
use minimum mean squared error (MMSE) channel estimation filters
that require knowledge of the specific Doppler frequency and the
specific signal-to-interference ratio (SIR) level. Unfortunately,
if the Doppler frequency or the SIR level changes, the filter also
changes. Thus, the RAKE receiver requires a Doppler estimator and
each finger of the RAKE receiver requires its own SIR
estimator.
Therefore, there is a need in the art for improved radio frequency
(RF) receivers. In particular, there is a need for improved channel
estimation filters for use in RAKE receivers. More particularly,
there is a need for a MMSE channel estimation filter that is not
specific to a particular Doppler frequency or SIR level provided
the SIR is maintained within a reasonable limit by use of downlink
power control.
SUMMARY OF THE INVENTION
The present invention comprises a low complexity channel estimation
filter for a DS-CDMA RAKE receiver that is optimized to work on a
range of Doppler frequencies using the average MMSE criterion. The
filter structure of the present invention remains the same at all
Doppler frequencies and SIR levels and is chosen to perform best on
the ensemble average of all Doppler frequencies.
Channel estimation is performed on each RAKE finger in order to
compensate for the complex channel gain that is associated with
each multipath in a scattering environment. In order to estimate
the channel gain, a known signal is required. The common pilot
channel (CPICH) is the phase reference for all common channels and
dedicated channels transmitted throughout the cell. The S-CPICH is
the phase reference for dedicated channels transmitted using beam
forming. The present invention uses whichever CPICH is the phase
reference to estimate the channel gain.
Multiplying the de-spread data symbols by the conjugate of the
channel gain can perform the channel compensation as well as
weighting the inputs for maximal ratio combining. Channel
estimation involves finding the mean of a non-stationary time
series of de-spread CPICH symbols. The present invention uses a
filter that effectively reduces the error in the estimate due to
additive noise, while having a low delay for following the mean. In
an advantageous embodiment, the present invention may be
implemented entirely using a digital signal processor (DSP).
The method for the channel estimation follows the following
steps:
1) Determine the ideal filter characteristics of a filter that
minimize the mean square error in the channel estimate. This is
done by assuming the Jake's spectrum for a particular Doppler
frequency, taking the channel estimation mean square error spectrum
at that Doppler, taking the mean of the mean squared error spectrum
over all Doppler frequencies, and minimizing the result over the
filter transfer function.
2) A discrete-time approximation of the obtained transfer function
is performed using the MMSE criterion in order to derive a
realizable filter at low complexity. According to an advantageous
embodiment of the present invention, such a filter may be
implemented as a cascade of a 3-tap finite impulse response (FIR)
filter and a single pole infinite impulse response (IIR) filter.
Such an embodiment is particularly useful because the operations
involved can be performed in one cycle of any DSP with 4
simultaneous MAC operations.
3) A fixed-point realization of the filter was obtained where the
filter coefficients as well as the data were 16-bit quantized. The
quantization levels are chosen to minimize the overall degradation
in the signal-to-noise ration (SNR) of the signal in overflow and
underflow conditions.
To address the above-discussed deficiencies of the prior art, it is
a primary object of the present invention to provide, for use in a
RAKE receiver, a channel estimator for determining channel
weighting coefficients for a finger of the RAKE receiver. According
to an advantageous embodiment, the channel estimator comprises: 1)
a first correlator for receiving a first pilot channel signal and
correlating the first pilot channel signal with a first pilot
channel chip pattern associated with the first pilot channel signal
to produce an output comprising a first pilot channel symbol
sequence; and 2) a first channel estimation filter capable of
receiving the first pilot channel symbol sequence and generating
first channel weighting coefficients, wherein the first channel
estimation filter minimizes a mean squared error in the first
channel weighting coefficients caused by additive noise and
variation in the channel (Doppler effects), wherein the mean
squared error is minimized across a range of Doppler frequencies
from 0 Hz up to a predetermined maximum Doppler frequency.
According to one embodiment of the present invention, the first
channel estimation filter comprises a cascade of a 3-tap finite
impulse response (FIR) filter and a single pole infinite impulse
response (IIR) filter.
According to another embodiment of the present invention, the
single pole infinite impulse response (IIR) filter is a Butterworth
filter having a maximally flat passband.
According to still another embodiment of the present invention, the
channel estimator further comprises: 3) a second correlator for
receiving a second pilot channel signal and correlating the second
pilot channel signal with a second pilot channel chip pattern
associated with the second pilot channel signal to produce an
output comprising a second pilot channel symbol sequence; and 4) a
second channel estimation filter capable of receiving the second
pilot channel symbol sequence and generating second channel
weighting coefficients, wherein the second channel estimation
filter minimizes a mean squared error in the second channel
weighting coefficients caused by additive noise and Doppler
effects, wherein the mean squared error is minimized across a range
of Doppler frequencies from 0 Hz up to a predetermined maximum
Doppler frequency.
According to a further embodiment of the present invention, the
second channel estimation filter comprises a cascade of a 3-tap
finite impulse response (FIR) filter and a single pole infinite
impulse response (IIR) filter.
According to a still further embodiment of the present invention,
the channel estimator as set forth in claim 5 wherein the single
pole infinite impulse response (IIR) filter is a Butterworth filter
having a maximally flat passband.
Before undertaking the DETAILED DESCRIPTION OF THE INVENTION below,
it may be advantageous to set forth definitions of certain words
and phrases used throughout this patent document: the terms
"include" and "comprise," as well as derivatives thereof, mean
inclusion without limitation; the term "or," is inclusive, meaning
and/or; the phrases "associated with" and "associated therewith,"
as well as derivatives thereof, may mean to include, be included
within, interconnect with, contain, be contained within, connect to
or with, couple to or with, be communicable with, cooperate with,
interleave, juxtapose, be proximate to, be bound to or with, have,
have a property of, or the like; and the term "controller" means
any device, system or part thereof that controls at least one
operation, such a device may be implemented in hardware, firmware
or software, or some combination of at least two of the same. It
should be noted that the functionality associated with any
particular controller may be centralized or distributed, whether
locally or remotely. Definitions for certain words and phrases are
provided throughout this patent document, those of ordinary skill
in the art should understand that in many, if not most instances,
such definitions apply to prior, as well as future uses of such
defined words and phrases.
BRIEF DESCRIPTION OF THE DRAWINGS
For a more complete understanding of the present invention and its
advantages, reference is now made to the following description
taken in conjunction with the accompanying drawings, in which like
reference numerals represent like parts:
FIG. 1 illustrates an exemplary wireless network in which mobile
station RAKE receivers using channel estimation techniques
according to the principles of the invention may be used.
FIG. 2 is a timing diagram illustrating the modulation pattern for
the common pilot channel (CPICH) signals in the wireless network in
FIG. 1 according to an exemplary embodiment of the present
invention;
FIG. 3 is a high-level block diagram of a RAKE receiver in an
exemplary mobile station according to one embodiment of the present
invention;
FIG. 4 is a block diagram of a channel estimator according to an
exemplary embodiment of the present invention;
FIG. 5 is a graph illustrating the power spectral density of the
received signal averaged over the entire range of Doppler
frequencies when the vehicle speed is a uniformly distributed
random variable (with maximum Doppler frequency of 500 Hz) and the
ideal channel estimation filter, G(f), in the wireless network;
FIG. 6 illustrates a channel estimation filter for calculating
weighting coefficients in a RAKE receiver according to an exemplary
embodiment of the present invention;
FIG. 7 illustrates a pole-zero plot for a channel estimation filter
in accordance with the principles of the present invention;
FIG. 8 illustrates a symbol de-rotator according to an exemplary
embodiment of the present invention.
DETAILED DESCRIPTION OF THE INVENTION
FIGS. 1 through 8, discussed below, and the various embodiments
used to describe the principles of the present invention in this
patent document are by way of illustration only and should not be
construed in any way to limit the scope of the invention. Those
skilled in the art will understand that the principles of the
present invention may be implemented in any suitably arranged
mobile station RAKE receiver.
FIG. 1 illustrates exemplary wireless network 100, in which mobile
station RAKE receivers using channel estimation techniques
according to the principles of the present invention may be used.
Wireless network 100 comprises a plurality of cell sites 121-123,
each containing a base station (BS), such as BS 101, BS 102, or BS
103. Base stations 101-103 communicate with a plurality of mobile
stations (MS) 111-114 over, for example, code division multiple
access (CDMA) channels. Mobile stations 111-114 may be any suitable
wireless devices, including conventional cellular radiotelephones,
Personal Communication Services (PCS) handset devices, personal
digital assistants, portable computers, or metering devices. The
present invention is not limited to mobile devices. Other types of
access terminals, including fixed wireless terminals, may be used.
However, for the sake of simplicity, only mobile stations are shown
and discussed hereafter.
Dotted lines show the approximate boundaries of the cell sites
121-123 in which base stations 101-103 are located. The cell sites
are shown approximately circular for the purposes of illustration
and explanation only. It should be clearly understood that the cell
sites may have other irregular shapes, depending on the cell
configuration selected and natural and man-made obstructions.
As is well known in the art, cell sites 121-123 are comprised of a
plurality of sectors (not shown), each sector being illuminated by
a directional antenna coupled to the base station. The embodiment
of FIG. 1 illustrates the base station in the center of the cell.
Alternate embodiments position the directional antennas in corners
of the sectors. The system of the present invention is not limited
to any particular cell site configuration.
In one embodiment of the present invention, BS 101, BS 102, and BS
103 comprise a base station controller (BSC) and one or more base
transceiver subsystem(s) (BTS). Base station controllers and base
transceiver subsystems are well known to those skilled in the art.
A base station controller is a device that manages wireless
communications resources, including the base transceiver stations,
for specified cells within a wireless communications network. A
base transceiver subsystem comprises the RF transceivers, antennas,
and other electrical equipment located in each cell site.
BS 101, BS 102 and BS 103 transfer voice and data signals between
each other and the public switched telephone network (PSTN) (not
shown) and the Internet via communication line 131, mobile
switching center (MSC) 140, and packet data serving node (PDSN)
150. MSC 140 is a switching device that provides services and
coordination between the subscribers in a wireless network and
external networks, such as the PSTN or Internet.
In the exemplary wireless network 100, MS 111 is located in cell
site 121 and is in communication with BS 101. MS 113 is located in
cell site 122 and is in communication with BS 102. MS 114 is
located in cell site 123 and is in communication with BS 103. MS
112 is also located close to the edge of cell site 123 and is
moving in the direction of cell site 123, as indicated by the
direction arrow proximate MS 112. At some point, as MS 112 moves
into cell site 123 and out of cell site 121, a hand-off will
occur.
The base stations may transmit from a single antenna or from two
antennas. If two antennas are used, the base stations may use
transmit diversity (e.g., space-time transmit diversity (STTD)) by
coding data in a space-time code and transmitting the pilot symbols
in an orthogonal pattern, such as the pattern illustrated in FIG.
2.
FIG. 2 illustrates timing diagram 200, which depicts the modulation
pattern for the common pilot channel (CPICH) signals in wireless
network 100 according to an exemplary embodiment of the present
invention. In FIG. 1, each of BS 101-BS 103 has two antennas that
may be used to communicate with MS 111-MS 114. Each of base
stations 101-103 may use a single antenna to communicate in a
non-transmission diversity (non-TD) mode with the mobile stations.
However, in an advantageous embodiment of the present invention,
each of base stations 101-103 may combat the effects of multipath
fading by transmitting from two antennas in a space-time transmit
diversity (STTD) mode.
In an exemplary embodiment, wireless network 100 is compatible with
the 3.sup.rd Generation Partnership Project (3GPP) standard. In a
3GPP system, during non-TD mode, a common pilot channel (CPICH)
signal is transmitted as a quadrature signal from a single antenna
using the pattern shown for Antenna 1 in FIG. 2, where A=1+j.
During STTD mode, a first common pilot channel (CPICH) signal is
transmitted as a first quadrature signal from a first antenna using
the pattern shown for Antenna 1 in FIG. 2, and a second common
pilot channel (CPICH) signal is transmitted as a second quadrature
signal from a second antenna using the pattern shown for Antenna 2
in FIG. 2.
FIG. 3 is a high-level block diagram of RAKE receiver 300 in
exemplary mobile station 111 according to one embodiment of the
present invention. RAKE receiver comprises antenna 301, radio
frequency (RF) front-end block 305, L fingers, including exemplary
fingers 310, 320 and 330, and combiner 340. Finger 310 comprises
delay element 311, multiplier 312, summer 313 and multiplier 314.
Finger 320 comprises delay element 321, multiplier 322, summer 323
and multiplier 324. Finger 330 comprises delay element 331,
multiplier 332, summer 333 and multiplier 334.
RF front-end block 305 downconverts the incoming RF signals
received from antenna 301 and produces a baseband or intermediate
frequency signal, which is sampled and quantized by an
analog-to-digital converter (ADC) to produce a sequence of digital
values, the signal R. The signal R is supplied as the input to each
of the L fingers. In each of the L fingers, there is a correlator
formed by a multiplier and a summer. For example, in finger 310,
the correlator is formed by multiplier 312 and summer 313, in
finger 320, the correlator is formed by multiplier 322 and summer
323, and in finger 330, the correlator is formed by multiplier 332
and summer 333.
In each finger, the signal R is initially delayed by some time
delay D(n) by the delay elements. The output of each delay element
is the input of the correlator for that finger. Thus, the
correlators are synchronized to each of the L strongest multipath
components by delaying the received signal R in each finger by an
appropriate amount of time D(n). The delayed samples of the
received signal R are then correlated with the chip pattern, c(k),
to produce a correlated output. The correlated outputs of the
correlators are then weighted by coefficients b(n) by the
multipliers 314, 324, and 334. Combiner 340 combines the weighted
outputs 340 and the resulting DATA OUT signal is the baseband
signal.
According to the principles of the present invention, the weighting
coefficients b(n) in each of the L fingers of RAKE receiver 300 are
calculated by a channel estimation filter that uses the pilot
channel signals transmitted by base stations 101, 102, and 103 and
that optimizes the weighting coefficients b(n) over a range of
Doppler frequencies using the average MMSE criterion. In an
advantageous embodiment of the present invention, a digital signal
processor (DSP) performs channel estimation.
FIG. 4 is a block diagram of channel estimator 400 according to an
exemplary embodiment of the present invention. There is one channel
estimator for each active finger. The inputs to channel estimator
400 are the integrated CPICH symbols and the CPICH symbol
pattern(s) for each antenna. The output(s) are the channel
estimates for each antenna-to-antenna path.
Channel estimator 400 comprises multipliers 410, 420A and 420B,
integrate and dump blocks 430A and 430B and channel estimation
filters 440A and 440B. It is noted that the elements in FIG. 4 may
be actual circuits in a fixed embodiment. However, if the RAKE
receiver is implemented in a digital signal processor (DSP), the
elements in FIG. 4 may be logical functional blocks, rather than
literal circuits.
Multiplier 410 receives pilot channel signal(s), CPICH, from a
pilot signal correlator. In non-TD mode, a single pilot channel
signal is received as a sequence of complex number symbols. In STTD
mode, two pilot channel signals are received as a sequence of
complex number symbols. Multiplier 410 multiplies the CPICH input
by the complex number (1+j)/2. The output of multipliers 410 is
sent to two correlators. The first correlator comprises multiplier
420A, integrate and dump block 430A, and channel estimation filter
440A. The first correlator produces the channel estimates, H1, a
complex number. The second correlator comprises multiplier 420B,
integrate and dump block 430B, and filter 440B. The second
correlator produces the channel estimates, H2, a complex number. H1
and H2 are used as the weighting coefficients (i.e., b(n)) for the
fingers of the RAKE receiver.
Multiplier 420A multiplies the complex symbol output of multiplier
410 by the pilot channel chip pattern for Antenna 1 (i.e., ANT. 1
CPICH PATTERN). The output of multiplier 420A is input to integrate
and dump block 430A, which integrates two symbols at a time (i.e.,
256 chips per symbol) and outputs (dumps) the integrated symbol
pairs (i.e., 512 chips) to channel estimation filter 440A.
In STTD mode, multiplier 420B multiplies the complex symbol output
of multiplier 410 by the pilot channel chip pattern for Antenna 2
(i.e., ANT. 2 CPICH PATTERN). The output of multiplier 420B is
input to integrate and dump block 430B, which integrates two
symbols at a time (i.e., 256 chips per symbol) and outputs (dumps)
the integrated symbol pairs (i.e., 512 chips) to channel estimation
filter 440B.
Channel estimation is required to be performed on each RAKE finger
in order to compensate for the complex channel gain associated with
each multipath in a scattering environment. In order to estimate
the channel gain, a known signal is required. The common pilot
channel P-CPICH is the phase reference for all common channels and
dedicated channels transmitted throughout the cells in wireless
network 100. The common pilot channel S-CPICH is the phase
reference for dedicated channels transmitted using beam forming. In
either event, whichever CPICH is present is used as the phase
reference for estimating the channel gain.
Multiplying the despreaded data symbols by the conjugate of the
channel gain can perform the channel compensation as well as
weighting the inputs for maximal ratio combining. Fundamentally,
the process of channel estimation is an estimation problem of
finding the mean of a nonstationary time series of despreaded CPICH
symbols. The objective is to use a filter that effectively reduces
the error in the estimate due to additive noise, while having a low
delay for following the mean value.
The starting point for constructing such a channel estimation
filter is to consider the ideal characteristics expected from a
channel estimation filter. The combination of the channel gains on
each of the RAKE fingers define the entire channel (almost
completely) as:
.function..times..function..times..times..function..times..delta..functio-
n. ##EQU00001##
After passing through this channel, the received and integrated
pilot channel symbols can be represented as:
r.sub.l(t)=A.sub.p(1+j).delta.(n-d.sub.l)*(h(n,k)+.eta.(n,k)).sub.l
The channel estimator derotates the original 45.degree. rotation in
the pilot symbols and filters the output with a real filter. This
operation may be represent by:
.times..function..times..times..times..function..function..times..delta..-
function..function. ##EQU00002##
The channel gain estimators act on the pilot signals demodulated by
each RAKE finger. The pilot signals are perturbed by noise. Thus,
the output of each channel estimator is:
(c.sub.r(k)+jc.sub.i(k)).sub.l=A.sub.pg(k)*.delta.(n-d.sub.l)*(h(n,k)+.et-
a.(n,k))
(c.sub.r(k)+jc.sub.i(k)).sub.l=A.sub.pg(k)*(c.sub.r(k)+jc.sub.i(k-
)+.eta.(k)).sub.l where A.sub.p is the pilot amplitude, r(n) is the
rake finger output and g(n) is the channel estimator.
The noise process perturbing the real and imaginary component of
the channel gain on each finger is Gaussian (or nearly so). Hence,
the optimum criterion for the channel gain estimators on each
finger is to minimize the mean square error (MSE) of the estimates.
e.sub.l(k)=[(c.sub.r(k)+jc.sub.i(k)).sub.l-(c.sub.r(k)+jc.sub.i(k)).sub.l-
]
By Parseval's theorem, minimizing MSE in the time domain is
equivalent to minimizing MSE in the frequency domain. In the
frequency domain, the following model may be considered:
H(f)=G(f)[H(f)+N(f)]
The problem is to find G(f) such that:
.function..times..times..times..function..times..intg..times..function..t-
imes..theta..times..function..theta..function..times..times.d.function..ti-
mes..times..times..theta..times..function..theta..function.
##EQU00003## which leads to:
.function..OMEGA..sigma..OMEGA..times..times..PHI..function..times..PHI..-
function..sigma. ##EQU00004##
In order to proceed further, it is necessary to assume a
probability density function (PDF) for Doppler frequency. It may be
assumed (for lack of any definitive data) that the velocity of the
mobile stations may be distributed uniformly from 0 kilometers per
hour (kmph) to 250 kmph. Then:
.PSI..function..times..PHI..function..intg..times..OMEGA..times..pi..time-
s..times..times..times. ##EQU00005##
This is plotted as in FIG. 5. FIG. 5 illustrates graph 500, which
depicts the power spectral density of the received signal averaged
in curve 505 over the entire range of Doppler frequencies when the
vehicle speed is an uniformly distributed random variable (with
maximum Doppler frequency of 500 Hz) and the ideal channel
estimation filter, G(f), in curve 510. As can be seen, the fall is
quite sharp at 500 Hz. In fact, there are no components beyond 500
Hz, which roughly corresponds to the maximum Doppler frequency at 2
GHz when the mobile speed is 250 kmph.
For a non-TD system, the sampling frequency is 15 kHz,
corresponding to integration of 256 pilot chips. Hence, a digital
filter needs to be constructed that closely approximates the G(f)
in FIG. 5 with a Nyquist frequency of 7.5 kHz, thereby satisfying
complexity constraints.
According to an advantageous embodiment of the present invention, a
second-order filter may be used. Since the overriding aim is to
have small attenuation in the stopband, an equiripple pass band
filter may be used. To keep the stopband attenuation minimum, an
equiripple 30 dB stopband and a 1 dB passband attenuation elliptic
filter may be used.
FIG. 6 illustrates channel estimation filter 440 according to an
exemplary embodiment of the present invention. A preferred digital
signal processor (DSP) can perform 4 MAC operations simultaneously.
A Butterworth filter involves 2 MAC operations, while a elliptic
(or any other 2.sup.nd order) filter involves 5 AMC operations. One
effective way to use the DSP to its fullest advantage is to use a
3-tap finite impulse response (FIR) filter followed by a
Butterworth stage as shown in FIG. 6.
Channel estimation filter 440 comprises a FIR filter stage and a
Butterworth filter stage (also known as maximally flat low pass
filter stage). The filter stage receives the input X(k). The FIR
filter stage comprises delay elements 610A and 610B, multipliers
620A, 620B and 620C, and adders 630A and 630B. The Butterworth
Filter stage comprises delay elements 650 and multiplier 660. Adder
640 adds the output of the FIR filter stage and the Butterworth
filter stage to produce the filtered output Y(k).
A first order approximation of the filter above will be a
Butterworth filter. The 3 dB cutoff frequency of the filter may be
chosen to match that of the ideal filter. This is around 300
Hz.
The corresponding filter has the form:
.function..times..times. ##EQU00006##
A first order filter will involve two multiplications and
additions. A first order filter may approximate the optimal filter
based on expected channel psd well, however, it produces a
significant attenuation (.about.6 dB @ 500 Hz) in the passband when
the highest Doppler is in use.
However, a cascade of a FIR filter and a Butterworth filter
overcomes the problems of a first order Butterworth filter by
itself. In the cascade design, the Butterworth filter design may be
kept identical to the one described above as it follows the
passband quite closely. The FIR stage may be designed to add
additional attenuation in the stopband while keeping very low
attenuation in the passband.
The equal tap FIR stage of: G.sub.F(z)=0.333(1+z.sup.-1+z.sup.-2)
gives an attenuation of less than 0.13 dB at 500 Hz and more than
10 dB from 3800 Hz and beyond.
Delaying the despreaded data by one sample prior to channel
compensation compensates for the extra sample of group delay caused
by the FIR stage. Thus, the combined filter is:
.function..times..times. ##EQU00007##
STTD Mode of Operation
The mobile station conducts a secondary search to determine whether
the base station is using the space-time transmit diversity (STTD)
mode of operation. Once it is determined that STTD mode is being
used, a STTD channel estimation technique is used.
In the STTD mode there are two channels to be estimated, the
channel from the first base station (BS) antenna (i.e., Antenna 1)
to the mobile station (MS), h.sub.1(t), and the channel from the
second BS antenna to the MS, h.sub.2(t). In STTD mode of operation
the two antennas transmit the pilot channel bit patterns in phase
and in anti-phase alternatively, as shown in FIG. 2. Hence, the
received and integrated CPICH symbols on each finger will follow
the pattern:
r.sub.l(t)=A.sub.p(1+j).delta.(n-d.sub.l)*(h.sub.l(n,k)+a.sub.nh.sub.2(n,-
k)+.eta.(n,k)).sub.l, where .alpha.a.sub.n.epsilon.{-1,1} forms the
phase sequence of the second antenna's CPICH transmission.
Hence, at any given time the effective received channel is the sum
or the difference of the two channels.
There are two methods for performing channel estimation:
1) Method 1: Estimate the sum and differences of the channel
separately by integrating separately over intervals when the CPICH
transmissions are in-phase and out-of-phase. The individual
channels can then be calculated by taking the sum and the
difference; and
2) Method 2: Integrate over intervals where the in-phase and
out-of-phase intervals are matched equally to produce
super-symbols. By choosing the sign of the despreading code, either
the first or the second channel may be estimated.
The advantages and disadvantages are of these methods are:
1) Method 1: The order of pilot symbols where the CPICH is from the
first and second antennas are the same (or different) signs is not
periodic. This means the first method requires housekeeping for
symbol counts. Also, because of this aperiodic characteristic, the
filter either must run at symbol rate or have special operations at
symbol boundaries.
2) Method 2: The sampling frequency gets halved leading to a 50%
MIPS or power savings. Also, no housekeeping beyond counting even
and odd symbols is necessary.
The second approach is preferred in this algorithm.
The operation involved is:
.times..function..times..times..times..function..function..times..times..-
times..times..delta..function..function..times..function..times..times..ti-
mes..times..delta..function..function..times. ##EQU00008## for
Antenna 1, and
.times..function..times..times..times..function..function..times..times..-
times..times..times..delta..function..function..times..function..times..ti-
mes..times..times..times..delta..function..function..times.
##EQU00009## for Antenna 2.
The resulting output from each channel estimator is:
(c.sub.sr(k)+jc.sub.sl(k)).sub.l=A.sub.pg(k)*.delta.(n-d.sub.l)*(h.sub.s(-
n,k)+.eta..sub.s(n,k))s=1,2
(c.sub.sr(k)+jc.sub.sl(k)).sub.l=A.sub.pg(k)*(c.sub.sr(k)+jc.sub.sl(k)+.e-
ta..sub.s(k)).sub.l
Note that due to averaging, the variance of the noise samples
.eta..sub.sl(k) is half of that of the non-TD case
.eta..sub.l(k).
The filter itself can take the same form as in FIG. 6. However,
because the filtering is taking place over supersymbols (which are
the sum of two consecutive symbols), the sampling frequency is
halved to 7.5 kHz. Hence, the filter coefficients must be adjusted
accordingly.
FIR-Butterworth Combination
The Butterworth filter in the non-TD design was designed to have a
3 dB cutoff frequency at 300 Hz. For a sampling frequency of 7500
Hz, the corresponding filter will be:
.function..times..times. ##EQU00010##
Note that both the pole position is closer to 0.5, leading to a
filter design with lesser vulnerability to quantization error. The
combined filter is obtained by cascading the FIR filter and the
Butterworth filter stages:
.function..times..times..times..times. ##EQU00011##
FIG. 7 illustrates pole-zero plot 700 for a channel estimation
filter in accordance with the principles of the present invention.
Note that the pole position is well inside the unit circle leading
to a filter design with more stability, better response time,
lesser group delay and lesser vulnerability to quantization
error.
Symbol De-Rotation in STTD Mode
In STTD mode, the transmitted symbols are transmitted from Antennas
1 and 2 of the base stations in the following fashion:
TABLE-US-00001 Transmission Time Antenna 1 symbol Antenna 2 Symbol
2m S.sub.2m -S*.sub.2m+1 2m + 1 S.sub.2m+1 S*.sub.2m
It is assumed that the first symbol interval in the frame was
symbol interval 0. Hence, the received symbols on even symbol
intervals are:
r.sub.2m=h.sub.1S.sub.2m-h.sub.2S*.sub.2m+1+.eta..sub.2m.
Similarly, the received symbols on odd symbol intervals are:
r.sub.2m+1=h.sub.1S.sub.2m+1+h.sub.2S*.sub.2m+.eta..sub.2m+1.
The optimal de-rotation for the received symbol on even symbol
interval is:
.times..times..times..times..times..times..times..times..eta..times..time-
s..eta..times. ##EQU00012## The optimal de-rotation on odd symbol
interval is:
.times..times..times..times..times..times..times..times..eta..times..time-
s..eta..times. ##EQU00013##
FIG. 8 illustrates symbol de-rotator 800 according to an exemplary
embodiment of the present invention. There is one symbol de-rotator
for every physical channel (other than CPICH) in every finger. The
inputs to the symbol de-rotator are the integrated traffic channel
(common or dedicated) symbols and the channel estimate(s) H1 and H2
from channel estimator 440. The outputs of symbol de-rotator 800
are the channel compensated symbols ready for combining. As in the
case of FIG. 4, it is noted that the elements in FIG. 8 may be
actual circuits in a fixed embodiment. However, if the RAKE
receiver is implemented in a digital signal processor (DSP), the
elements in FIG. 8 may be logical functional blocks, rather than
literal circuits.
Symbol de-rotator 800 has a Non-TD Output and a STTD Output. In
non-TD mode, each channel estimate, H1, from channel estimator 400
is complex conjugated by complex conjugate block 810A and the
output is applied to one input of multiplier 815. The other input
of multiplier 815 receives the traffic channel symbols from traffic
channel correlator 890. The output of multiplier 815 is the channel
compensated symbols that form the Non-TD Output.
In STTD mode, complex conjugate block 810A and multiplier 815
operate as in non-TD mode. The Non-TD output is applied to a first
input of summer 860.
The traffic channel symbols from traffic channel correlator 890 are
complex conjugated by complex conjugate block 810B and the output
is applied to serial-to-parallel (S>P) block 820, which converts
the symbol data from serial to parallel. Alternating symbols from
S>P block 820 are stored in registers 825 and 830. Even symbols
(i.e., symbol 2m) are stored in register 825 and odd symbols are
stored in register 830 (i.e., symbol 2m+1). The symbol data in
register 830 is transferred to register 840, but the symbol data in
register 825 is negated by inverter 835 and then stored in register
845. The symbol data in registers 840 and 845 are then read by
parallel-to-serial (P>S) block 850.
The serial symbol data from P>S block 850 is multiplied by the
channel estimate, H2, from channel estimator 440 by multiplier 855.
The channel compensated symbols from multiplier 855 are then
combined with the channel compensated symbols from the Non-TD
Output to form the STTD Output.
The prior art includes a Wiener filter-based MMSE channel
estimation that requires knowledge of SIR and Doppler. Hence it
requires a Doppler and SIR estimator. The filter structure will
change dynamically when these quantities change. The present
invention overcomes numerous disadvantages of the prior art
including:
1) There is no need for a Doppler estimator or a per finger SIR
estimator;
2) The filter structure does not change with changes in is Doppler
and SIR; and
3) There is not need for time-intensive calculation, such as matrix
inversions.
These improvements make the design simpler, consume less power,
require less silicon area, and the like.
The present invention performs sub-optimally compared to the prior
art at a particular Doppler value and SIR setting. However, the
present invention gives the best performance for the ensemble
average for all Doppler settings and performance simulations
demonstrate acceptable performance at the entire range of expected
Doppler frequencies. The SIR level is typically maintained constant
under closed-loop power control and hence this advantage of the
related art is of little practical value.
Although the present invention has been described with several
embodiments, various changes and modifications may be suggested to
one skilled in the art. It is intended that the present invention
encompass such changes and modifications as fall within the scope
of the appended claims.
* * * * *